Software Daily Summary
Software Daily Summary: What Happened on October 31, 2026
Heng Li urges editors to reject suspected AI-written reviews
The researcher says AI reviews can burden authors with weak or excessive feedback, but their effect in a standard response process has not been tested.
Today in brief
Heng Li argues that editors should be able to discard peer reviews they suspect were fully generated by AI, citing weak judgment, nitpicking and excessive length.1
AI reviews and author responses
Li's concern centers on review processes that require authors to answer every reviewer comment, including AI-generated ones.1 He says one study found authors rated human reviews more highly than AI reviews on most measures, while another surveyed authors after AI systems reviewed 23,000 AAAI-26 submissions.1 In that survey, authors preferred AI reviews in more categories but cited weak overall judgment, nitpicking and excessive length as leading flaws.1
Li says the AAAI-26 study did not test AI reviews in a typical process where authors respond point by point, and that he found no study assessing them in a realistic standard review cycle.1 He recommends that editors discard reviews they detect or strongly suspect were fully AI-generated.1